- Monsieur Guillaume PLANTIN , Chercheur associé TSE, Professeur Sciences Politiques Paris
- Monsieur Milo BIANCHI, Chercheur TSE, Professeur, Université Toulouse 1 Capitole
- Monsieur Johan HOMBERT, Professeur associé HEC Paris
- Monsieur Pierre FLECKINGER, Professeur, Université Paris 1 Panthéon-Sorbonne
Résumé (enanglais):
This thesis contains three chapters that study the behaviors of three types of financial intermediaries: fund managers, banks and financial analysts.
The first chapter investigates how fund managers' labor market competition affect their investment strategies. In an economy in which asset management firms compete both for money to manage and for managers to manage it, I show that a manager's compensation scheme is endogenously benchmarked against the market return when the expected market return is high whereas it is benchmarked against the performance of other active managers when the expected market return is low. Managers concerned about labor market outcomes adjust their investment strategies accordingly: when they expect high returns, managers correlate their investments with one another so as to accelerate the revelation of their abilities; when anticipating low returns managers differentiate their investments so as to increase the relative returns that they can generate compared to their peers. I then test the model's predictions empirically. I build an individual-level measure of managers' beliefs on market trends using sentiment analysis and exploit the measurement to show that managers' portfolios are more correlated when they are more optimistic about future market returns. The effect is more pronounced for managers with less work experience among their peers and thus more important career concerns. Additionally, managers' investments are more correlated when there are more fund entries and larger capital inflow rates, which indicate a less competitive labor market for managers.
The second chapter develops an endogenous network model of banks to study how financial networks interact with financial regulation. In the model, to meet regulation requirements banks optimally choose the number of counterparties to connect with and the amount of shares to trade with these counterparties, facing the trade-off between reserving costly capital and holding more diversified (and thus less profitable) balance sheets. I firstly solve the network structures for given regulation requirements, and derive a partial order on these networks based on their systemic riskiness. Then the equilibrium regulation is discussed. I show that: 1) When the regulator has commitment power, the first best outcome can be achieved; 2) When the banks have commitment power, banks commit to a bang-bang network formation pattern: to form the systemically least-risky networks if the regulation is looser than a threshold and the most-risky networks if the regulation is tighter than the threshold; being afraid that banks may form the most-risky networks, the regulator imposes looser regulation requirements than the socially optimal one; the least-risky networks are then formed but the maximal social welfare of the most-risky networks realizes.
In the third chapter (jointly with Kun Li), we empirically identifies herding among financial analysts. This chapter is motivated by the fact that empirical works on herding usually focus on behavior patterns correlated across individuals, but such behavior could well be the results of correlated information arrival to independently acting agents. We explore daily analysts' forecasts and exploit the All-American Research Team award as a reputation shock to identify herding among them. We find that after award the distance between non-winning analysts is reduced more on stocks that are covered by a first time winner. A detailed investigation into the direction of moves shows that when first time winners moves first, they move away from pre-award forecasts of the non-winning analysts and when non-winning analysts move first, they move closer to pre-award forecasts of the first time winners. We further show that non-winning analysts herd more on stocks for which the winning analysts are more accurate or non-winning analysts are less accurate, which suggests that analysts herd rationally.
